Big Tech’s Revenue Surge Is Not Reaching the Bottom Line

Four of the largest companies in the S&P 500 are set to report earnings on the same afternoon, and each is expected to post double-digit revenue growth. Yet their earnings-per-share figures tell a different story: nearly flat across the group, and in one case, an outright decline. That gap between top-line momentum and bottom-line compression is the defining tension of this earnings season, and it traces directly to how Alphabet, Meta, Amazon, and Microsoft are choosing to spend.

On 29 April 2026, these four companies report after market close in what amounts to the most consequential single session in the quarterly earnings calendar. Markets have recovered to record highs following a volatile start to the year shaped by AI equity rotation, elevated oil prices, and geopolitical uncertainty. Investors arrive at this week already primed to interrogate the relationship between AI capital expenditure and financial returns.

What follows is a framework for understanding why strong revenue growth is not translating into proportional shareholder returns, what each company’s spending strategy signals about its competitive positioning, and what the long-term payoff thesis actually rests on.

The paradox at the heart of Big Tech’s Q1 2026 results

The numbers, placed side by side, surface the question before any analysis needs to.

Combined, these four names represent more than 18% of the S&P 500 by weighting. All four are growing revenue at double-digit rates. Three of the four are expected to deliver EPS growth in the low single digits or worse.

Alphabet offers the sharpest illustration: revenue up 19%, earnings per share down 5%. The top line is accelerating while the bottom line contracts.

This is not a failure of execution. It is a structural condition, the direct financial consequence of capital expenditure decisions made deliberately at the executive level. The divergence is not isolated to one business model or one competitive dynamic. It runs across the entire cohort, and that is what makes it analytically significant. Microsoft, projecting 17% EPS growth, stands as the outlier, a data point worth interrogating separately.

The broader concern underpinning all four companies’ spending decisions is AI capital expenditure ROI uncertainty: consensus hyperscaler capex has already reached hundreds of billions of dollars, yet no major provider has announced a credible path to proportionate returns at the scale of investment currently being made.

What $635 billion in AI infrastructure actually looks like

The combined AI infrastructure spending projected for these four companies in 2026 falls in the range of $635-665 billion. That figure is abstract at headline level. At the company level, it becomes concrete.

The physical assets absorbing this capital share common categories across all four companies:

• Custom AI chips (Google’s TPUs, Amazon’s Trainium and Inferentia processors)
• GPU clusters at data centre scale
• Data centre real estate, including new builds and expansions
• Fibre and power infrastructure to support computing density

Meta’s spending escalation is the steepest in percentage terms: $115-135 billion for 2026, up from $72 billion in 2025, representing a 60-88% increase. CEO Andy Jassy at Amazon has noted that customer commitments cover nearly the entirety of the company’s planned $200 billion capex, a framing designed to pre-empt the overbuilding concern. Alphabet has locked in a new five-year TPU development agreement with Broadcom, signalling that the spending horizon extends well beyond a single fiscal year.

These are not speculative allocations. Management teams across the group have framed the spending as demand-driven, characterising their infrastructure positions as behind capacity rather than ahead of it.

Why revenue and earnings are moving in opposite directions

How capex hits the income statement

The mechanics that produce the revenue-earnings divergence are structural, not mysterious. Capital expenditure flows through the income statement in three stages:

1. The company spends cash to acquire or build a long-lived asset (a data centre, a GPU cluster, a custom chip fabrication agreement).
2. That spending is capitalised on the balance sheet as a fixed asset rather than expensed immediately.
3. The asset is depreciated over its useful life, typically three to five years for computing equipment, with the depreciation charge running through the income statement each quarter.

The result: today’s record capex does not fully hit this quarter’s earnings. Instead, it creates a rolling depreciation burden that accumulates as each successive investment cycle layers onto the last. Elevated depreciation from prior spending cycles is now running through the profit and loss simultaneously with new commitments, compounding the drag on reported earnings even as revenue accelerates.

Meta CFO Susan Li noted on the 28 January 2026 earnings call that compute demand exceeds supply, with capacity constraints expected through much of 2026. GPUs, she indicated, are generating advertising revenue within weeks of deployment.

Where AI monetisation is already showing up in the numbers

The evidence that spending is translating into revenue is uneven but real. Amazon’s AI chip division has reached an annualised revenue run rate of $50 billion, growing at more than 100% per year according to CEO Jassy’s commentary. Google Cloud grew 48% year on year to $17.7 billion in Q4 2025, providing a benchmark for cloud monetisation velocity. Microsoft Cloud revenue reached $51.5 billion in fiscal Q2 2026.

Microsoft is the cohort outlier in a meaningful sense: its projected 17% EPS growth suggests that Copilot monetisation and Azure AI services are already flowing through the income statement more efficiently than at peers. Bloomberg reported that Copilot paid conversion met ambitious internal benchmarks last quarter.

Where the returns remain forward-looking

Not all of the spending maps to near-term AI monetisation. Alphabet’s potential $40 billion additional investment in Anthropic (in which it holds a 14% ownership stake) represents a bet whose payoff is entirely future-dated. Amazon’s $10.8 billion acquisition of Globalstar targets a satellite service with a projected 2028 launch, illustrating that portions of the capex envelope serve strategic objectives beyond this year’s AI revenue.

How each company is telling a different version of the same story

The four companies share a common structural dynamic, but the risk profiles and monetisation timelines diverge in ways that matter for anyone evaluating individual positions.

• Microsoft is the near-term monetisation leader. Copilot paid conversion is reportedly performing ahead of internal expectations, and the company restructured its OpenAI relationship (introducing a payment cap and non-exclusivity provisions) in a move that signals more disciplined return management. EPS growth of 17% separates it from the rest of the cohort.
• Meta is running the highest revenue growth (31%) against the steepest spending escalation. A 10% headcount reduction (approximately 8,000 employees) partially offsets the capex ramp. The company launched its Muse Spark AI model in April 2026, and CFO Li’s commentary positions the company as capacity-constrained rather than speculatively overbuilt.
• Alphabet carries the most visible EPS decline risk. The projected 5% year-on-year earnings contraction reflects the weight of the Anthropic investment commitment and the Broadcom TPU partnership. Revenue growth of 19% remains strong, but the earnings compression is the sharpest in the group.
• Amazon has been the most direct in pre-empting investor scepticism. CEO Jassy’s annual letter characterised the market’s negative stock reaction to the $200 billion capex announcement as an overreaction, citing customer commitment coverage as evidence that the spending is demand-backed, not speculative.

The structural difference in business models matters too. Meta and Alphabet are primarily monetising AI through advertising yield improvement: better targeting, higher click-through rates, and more efficient ad delivery. Amazon and Microsoft are monetising through cloud infrastructure rental and enterprise software subscriptions, a model where utilisation rates and seat counts determine the speed of returns.

Understanding the AI infrastructure investment cycle

Large-scale technology infrastructure investments have historically followed a recognisable multi-year pattern. The current AI spending cycle fits this framework, though the pace appears to be compressing.

The cycle moves through three phases:

1. Investment and margin compression. Heavy upfront capital expenditure dilutes margins and depresses earnings relative to revenue growth. This is where the cohort currently sits.
2. Utilisation ramp. As infrastructure comes online and workloads scale, fixed costs are spread across a growing revenue base. Margin pressure begins to ease.
3. Operating leverage and earnings acceleration. Once infrastructure costs are sunk, incremental revenue flows through at substantially higher margins, and earnings growth catches or exceeds revenue growth.

The parallel to the cloud buildout of the early 2010s is instructive without being exact. AWS spent years as a margin-dilutive segment within Amazon before becoming the company’s most profitable division. The current AI capex cycle appears to be moving faster: Meta’s disclosure that GPUs generate advertising revenue within weeks of deployment compresses the timeline between phases one and two relative to prior infrastructure buildouts. Amazon’s $50 billion AI chip run rate, growing at more than 100% year on year, suggests utilisation is already ramping at scale.

Shai Luft of Bench Media has framed the spending calculus this way: current AI investment represents roughly 15-20% of revenue, and the risk of under-investing, given the potential disruption to core search, cloud, and advertising businesses, may exceed the risk of over-investing.

The question for investors is not whether the cycle will play out, but where in the cycle each company currently sits, and how quickly phase two transitions into phase three.

Societe Generale has warned that the investment cycle carries a cash flow consequence most investors have not fully priced: hyperscaler free cash flow turning negative by late 2026 before recovering in Q1 2027, a trajectory that would put further pressure on reported earnings even as AI revenue begins to ramp.

What investors should watch when results drop on April 29

What the numbers need to show

Four categories of data will determine whether the AI investment thesis is tracking or under pressure:

• Capex guidance reaffirmation. Any company pulling back on its 2026 spending commitment would signal a shift in competitive conviction. Any increase would suggest management is seeing demand that has not yet been fully disclosed.
• AI monetisation metrics. Cloud growth rates for Google Cloud (benchmarked at 48% year-on-year growth in Q4 2025), AWS, and Azure will signal whether enterprise AI adoption is accelerating. Any new AI-attributed revenue disclosures will be closely scrutinised.
• Cloud growth trajectories. Microsoft Cloud’s $51.5 billion fiscal Q2 2026 figure sets the Azure context. Any deceleration across the three cloud platforms would weigh on the broader thesis.
• EPS surprises. Given the compressed expectations, even modest beats relative to the flat projections could shift market sentiment.

What management commentary needs to say

Beyond the numbers, guidance language carries weight. Investors should monitor whether companies maintain, raise, or hedge their capex commitments in light of tariff policy developments and broader market volatility during Q1 2026.

Any new AI revenue run-rate disclosures, following Amazon’s precedent of quantifying its AI chip business at $50 billion annualised, would provide the kind of concrete monetisation evidence the market is looking for. Management tone on demand conditions relative to supply constraints, particularly whether capacity remains the binding limitation, will indicate whether the spending trajectory is intact.

The S&P 500 and Nasdaq sit at record highs as of late April 2026. Expectations are elevated, which raises the bar for a positive market reaction even if headline numbers land in line with consensus.

JPMorgan has explicitly warned investors about options market implied moves this earnings week, with the signal reflecting expected magnitude of price swings rather than direction; the VIX closed at 19.31 on 23 April 2026, and a geopolitical risk premium tied to oil price pressure is amplifying those moves beyond what earnings uncertainty alone would generate.

Where the analysis lands

The tension between record revenue growth and compressed earnings across Alphabet, Meta, Amazon, and Microsoft is not a sign of strategic failure. It is the visible cost of an infrastructure bet that management teams at all four companies have concluded they cannot afford not to make. The alternative, ceding AI infrastructure dominance to a competitor, is assessed as the more dangerous outcome.

The genuine uncertainty remains: the investment thesis rests on utilisation ramps and operating leverage that have not yet fully materialised in reported numbers. Microsoft, with its projected 17% EPS growth, offers the clearest current evidence that the cycle can compress and deliver earnings simultaneously. The other three are at earlier stages of the same curve.

29 April will not resolve the question of whether $635-665 billion in combined AI infrastructure spending was the right call. But it will provide the first 2026 data points on monetisation velocity, capex conviction, and management confidence in the demand trajectory, all inputs to a multi-year thesis playing out in real time.

The metrics that matter most are those that speak to where each company sits in the infrastructure investment cycle, not the headline EPS number that has already been telegraphed as subdued.

For investors who want to stress-test the bullish infrastructure thesis further, our deep-dive into Hussman’s debt-dependency argument for AI valuations examines how over $38 trillion in federal debt and more than $2.5 trillion in corporate debt maturing before 2027 could compress AI sector margins independently of whether the infrastructure investment itself proves commercially sound.

This article is for informational purposes only and should not be considered financial advice. Investors should conduct their own research and consult with financial professionals before making investment decisions. Past performance does not guarantee future results. Financial projections are subject to market conditions and various risk factors.